MT Højgaard Q2 2026 slides: tunnel hit masks growth in order book

MT Højgaard reported a Q2 operating loss of DKK 77m driven by a DKK 175m Nordhavn Tunnel write-down, despite revenue of DKK 2,505m (down 6% YoY). It maintained full-year 2026 guidance of revenue DKK 10.0–10.5bn and EBIT DKK 225–275m, supported by 91% order coverage, and grew its total order portfolio to a record DKK 24.7bn (+22% YoY). Cash flow was pressured (operating cash flow DKK -69m vs +115m prior year) due to working-capital increases, and management flagged that guidance for Enemærke & Petersen depends on an expected land sale in 2H.

Analysis

The key mechanism is not just that AI demand is still present; it is that the next leg of the cycle is shifting from proof-of-concept to sustained capacity buildout. That matters because it extends pricing power, keeps utilization tight across the GPU stack, and supports a longer runway for estimate revisions across the semiconductor complex. The second-order winners are the bottlenecked suppliers and infrastructure enablers — TSM, AVGO, ANET, and high-end server assemblers — while lower-quality AI beneficiaries with no direct scarcity advantage are likely to lag as capital rotates toward the actual toll collectors.

Near term, the main risk is not demand collapse but digestion: if hyperscaler capex growth normalizes faster than consensus expects, NVDA can still beat while the stock multiple compresses. The biggest falsifier over 1-3 months would be any sign that gross margin expansion stalls, lead times shorten materially, or management starts talking about a more balanced supply/demand profile rather than scarcity. Over 6-18 months, export controls and a possible step-down in incremental training spend are the real tail risks; if those do not materialize, NVDA remains one of the few large-cap names with both top-line growth and operating leverage.

Contrarian view: the market may be underestimating how long the AI buildout stays capex-intensive, but overestimating how evenly that spending will accrue across the AI basket. The strongest risk/reward is still in the leader with pricing power, not the second-tier names trading on narrative momentum. If the stock gaps hard on the print, the trade is to wait for volatility to normalize rather than chase; the better entry is usually after the first post-earnings digestion day rather than at the open.

Key Decisions for Investors

• Buy NVDA on a 3-5% post-print pullback or after first-day volatility settles; target 8-12% upside over 1-3 months if guidance and margin commentary stay firm.
• If already long the AI basket, rotate toward NVDA and out of weaker AI proxies; pair long NVDA / short AMD over 1-3 months to express winner-takes-more pricing power and share gains.
• Add TSM and AVGO as second-order beneficiaries if management commentary implies continued supply-chain tightness; these names should benefit if the capex cycle broadens beyond GPUs.
• Use call spreads rather than outright stock if implied volatility remains elevated; prefer 1-3 month maturities to capture the estimate-revision window while limiting downside if the market fades the beat.
• Set a thesis-falsifier alert on any guide-down in gross margin, a commentary shift to demand normalization, or a material deceleration in hyperscaler capex in the next earnings cycle.